AI/ML CASE STUDY
Tackling a $35M+ Attrition Challenge with
Machine Learning
johnson controls and beyond the arc
Johnson Controls International (JCI), founded in 1885, is an industry leader in fire, security, and HVAC technology and services for buildings. After successfully using AI to deliver actionable insights for their building solutions, JCI saw an opportunity to optimize internal operations and reduce churn.
Is your company struggling to reduce churn rate? Find out how Beyond the Arc can help with churn prediction with machine learning.
Every 1% reduction in attrition could help JCI retain $35M in revenues.
— VP & President of Global Services,
Johnson Controls
JCI’S BUSINESS CHALLENGE
How could JCI reduce churn using AI/ML
– with no usable data sets and no data scientists?
– with no usable data sets and no data scientists?
Improving the attrition rate would have a significant financial impact. In an Investor Day presentation, JCI’s VP & President of Global Services noted that every 1% reduction in the attrition rate would impact $35M in revenues.
In 2021, the company added a new focus on AI to better understand customer attrition and generate predictive insights that could help reduce churn. To lead this charge, Terry Miller, Executive Director – Global Advanced Analytics, joined Johnson Controls. Terry has spent over 10 years working with OEMs to evaluate and optimize industrial processes with the help of statistical models and ML for predictive maintenance and process optimization.
It was a colossal challenge. Terry had no usable data sets, and no data science team or data engineers. In short order, he had to build a global data team with new AI/ML capabilities to help improve business outcomes on a major scale.
Given the urgent nature of the project, JCI partnered with Beyond the Arc, a leader in AI/ML analytics consulting with deep experience in leading enterprise projects like this one.
Working closely with JCI, Beyond the Arc deployed their AI development process. It includes phases for discovery, data preparation, model selection and iterations, data visualization, rapid prototyping and deployment. With this systematic approach the team was able to develop an AI solution using machine learning for customer churn prediction.
Data challenges in churn prediction with machine learning
Next, Beyond the Arc led Terry’s team into the technical phases of the project to predict churn. The biggest challenge in trying to predict churn was the data.
Making sense of the data was a monumental task. “It was very laborious and very time-consuming, and Beyond the Arc was instrumental in helping us with that,” Terry adds. Through deep dive interviews with each business, the project team was able to:
- match data sources to established key inputs
- map where and how each business captures that data
- understand exactly what the data means for each line of business
With Beyond the Arc, we had prototype models in 90 days. It feels like a Herculean feat.
— Terry Miller, Executive Director
Global Advanced Analytics, Johnson Controls
Scaling up machine learning solutions to reduce churn
Partnering with Beyond the Arc, Terry and his team identified the success factors for solving a customer churn prediction problem using machine learning at JCI. It equipped them to clearly define the next problem and solve it to attain the desired outcome.
Business value and accomplishments of using machine learning to predict churn
Taking action with ML-driven predictive insights is paying off. Terry highlights that JCI has had, “pretty significant successes.”
How does that success translate to measurable savings by reducing attrition?
By testing and verifying results in multiple ways with AI/ML, the company is identifying $100 million a year annualized of protectable revenue. While that savings is not exclusively a result of churn prediction, integrating AI/ML for this purpose has “mattered significantly” in terms of protected revenue they can track.
As a global enterprise, JCI is only at the beginning of their AI/ML journey for internal process optimization but making rapid progress. A machine learning customer churn reduction project received the company’s Chairman’s Award. Partnering with Beyond the Arc has given them a strong foundation. With an effective AI development process, the company is equipped with best practices for unlocking value from complex data. Tackling churn prediction with machine learning is only one of the business problems they now have on the radar.
With AI/ML we’re identifying $100M a year annualized of protectable revenue.
— Terry Miller, Johnson Controls